JP2010277480A - Method and device for search of important person, and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To search for a person important for an arbitrary person in consideration of indirect human relationships and with a smaller calculation amount for the number of persons to be treated. <P>SOLUTION: An important person search device 100 searches a person important for the arbitrary person by human relationship information showing whether two arbitrary persons have human relationships and hobby information showing hobbies of respective persons with respect to a plurality of persons. A center ID human rank calculation unit 200 uses a prescribed recurrence formula (formula (1)) expressing a Markov process to calculate human ranks representing degree of importance for each of other persons for a central person based on the human relationship information. A center ID hobby vector storage unit 300 generates hobby vectors from the hobby information. A center ID human rank-hobby coupling score calculation unit 400 applies human ranks and hobby vectors to a prescribed human relationship score calculation formula (formula (2)) to calculate human relationship scores of the other persons for the central person. <P>COPYRIGHT: (C)2011,JPO&INPIT

Description

  The present invention relates to a technique for searching for an important person for an arbitrary person based on information on a plurality of persons.

  In recent years, there has been a demand for improvement in technology for searching for an important person for an arbitrary person based on information on a plurality of persons. Here, for an arbitrary person, an “important person” means a person who satisfies at least one condition of “a person with a close relationship” and “a person with a high possibility of having a common hobby”.

  For example, such a search technique is used as a means for grasping the degree of human relationship between another person and oneself when communicating with others on the network like a social network service (SNS) on the Internet. Useful. Such a search technique is also useful when it is desired to grasp the importance of information issued by others in an information search service on the Internet.

  However, in the service via the network such as the Internet, it is difficult to grasp the human relationship on the network, the degree of friendship (human relationship) between a person (subject) and a person (other) (distance), It has been difficult to grasp the importance level of a certain person (other person), the importance level of information issued by a certain person (other person), and the like. In order to solve this problem, there have been the following conventional techniques.

For example, Non-Patent Document 1 discloses a technique for calculating and displaying a direct relationship (number of exchanges) with a certain person (other person) as the importance of the other person for the subject.
Non-Patent Document 2 discloses a technique for grasping the importance of a web page and the reliability of the web page using the link structure of the web page in the field of web pages on the Internet.

Y. Matsuo, J. Mori, and M. Hamasaki. “POLYPHONET: An Advanced Social Network Extraction System from the Web”, WWW 2006, (Scotland), ACM, 2006 May 23-26, p.397-406 L. Page, S. Brin, R. Motwani, T. Winograd, “The PageRank Citation Ranking: Bringing Order to the Web”, Tech. Rep., (USA), Stanford University, 1998 January 29, p.1-17

  However, the conventional methods for grasping human relationships on the network have the following problems.

  First, in the technology of Non-Patent Document 1, when grasping a human relationship, a direct relationship between a person (subject) who wants to know the human relationship and another person (a friend list of the subject, from the subject to the other's diary) Only information on visit history / comments, common hobbies, etc.) is used. Therefore, it was difficult to consider indirect friendships such as friends of friends.

  In addition, in the importance level grasping method using the link structure disclosed in Non-Patent Document 2, as the number of persons increases, a graph expressing human relations (mathematically, an adjacency matrix (node (person) and edge (relationship) The matrix representing the graph etc.)) becomes enormous and the amount of calculation becomes enormous. For this reason, the importance of each person is different depending on the subject (hereinafter referred to as “personalized”), such as human relations, and it is not suitable for the purpose of performing rank calculation.

  The present invention has been made in view of the above-described problems, and it is possible to search for an important person for an arbitrary person with a small amount of calculation in consideration of indirect human relations and the number of persons to be handled. Is an issue.

  In order to solve the above-described problem, the present invention relates to a plurality of persons from any person relationship information indicating the presence or absence of a person relationship between any two persons and each person's hobby information indicating a hobby. An important person search device for searching for an important person, and a storage unit for storing human relationship information and hobby information, and the human relationship information and hobby information stored in the storage unit, When a human relationship score indicating the importance of a person is calculated and the human relationship score calculation unit that stores the human relationship score in the storage unit and input of identification information of an arbitrary person is accepted, A human relationship score acquisition unit that acquires a human relationship score of another person and a result display unit that displays the human relationship score acquired by the human relationship score acquisition unit, and a human relationship score calculation unit A person who has a predetermined person as a central person and refers to the human relation information stored in the storage unit and who has a human relation with the central person and the central person directly or indirectly through one or more persons is “0”. And a person set matrix that represents other persons as “0” and creates a person set matrix between the central person and each person who has a human relationship directly with the central person or indirectly through one or more persons. A predetermined initial vector is created for a human rank vector having a human rank indicating the importance for the central person based on the human relation information as a component, and the central person and the central person are directly or via one or more persons Create a personalized vector whose component is the ratio of trustworthiness from the central person with each person who has an indirectly human relationship, and trust the person other than the central person from the central person. The first coefficient indicating the degree is multiplied by the product of the person set matrix and the human rank vector, the second coefficient indicating the reliability of the central person from the central person is multiplied by the personalized vector, and the multiplication is performed. Apply a person set matrix, an initial vector of human rank vectors, and a personalized vector to a predetermined recurrence formula based on a Markov process that updates a human rank vector by adding two vectors. By calculating the number of times, the human rank calculation unit that calculates the human rank vector having the human rank as a constituent element and the hobby information stored in the storage unit are referred to, and the hobby with each other person is related to the central person. Common hobby information that calculates the common number and indicates the number of hobbies related to the central person and the number of common hobbies for each person who has a common hobby with the central person The common hobby information creation unit to be created, and for each other person, the human rank of the other person in the calculated human rank vector is multiplied by the third coefficient indicating the importance level of the human rank, and the hobby commonness The fourth coefficient indicating the degree of importance is calculated based on the common hobby information, and the hobby commonality between the central person and the central person directly or indirectly through one or more persons and the whole person having a human relationship Calculated by applying human rank and common hobby information to a given human relationship score calculation formula that multiplies the value indicating the degree and adds the two multiplied values. Thus, a human rank hobby combining unit that calculates a human relation score of the other person with respect to the central person is provided.

  According to this invention, the above-described person set matrix, initial vector, and personalized vector are created from the human relationship information, and applied to the predetermined recurrence formula representing the Markov process, so that the human relationship information is obtained. Calculate a human rank indicating the importance of each other person for the central person based. Also, the hobby vector described above is created from the hobby information. Then, by applying the human rank and the hobby vector to a predetermined human relationship score calculation formula, the human relationship score of another person with respect to the central character is calculated. In this way, when calculating the human relationship score, a matrix with a small number of dimensions created from information about a person who has a human relationship directly with one or more persons directly or indirectly with the central person and the central person, rather than all the persons, By using a vector for calculation, it is possible to search for an important person for an arbitrary person with a small amount of calculation for the number of persons handled.

In the present invention, the person set matrix is preferably an adjacency matrix or a transition matrix.
According to this invention, the person set matrix can be expressed using an adjacency matrix or transition matrix which is an existing mathematical method, and the implementation of the present invention becomes easy.

In the present invention, it is preferable that the human relationship score calculation unit applies a predetermined number from the larger number of common hobbies in the hobby vector to the predetermined human relationship score calculation formula.
According to this invention, the calculation cost can be further reduced.

Further, the present invention is a program for causing a computer to function as the important person search apparatus described above.
According to this invention, a computer in which this program is installed can realize each function based on this program.

  According to the present invention, it is possible to search for an important person for an arbitrary person in consideration of indirect human relations and with a small amount of calculation for the number of persons to be handled.

It is a block diagram of the important person search apparatus of this embodiment. It is a flowchart which shows the process by an important person search apparatus. It is a block diagram of a human relation score calculation part and its periphery. It is a block diagram of a center ID human rank calculation part and its periphery. It is a block diagram of a human rank calculation part and its periphery. It is a block diagram of the neighborhood friend matrix preparation part and its periphery. It is a block diagram of a center ID hobby vector creation unit and its periphery. It is a block diagram of a hobby vector creation part and its periphery. It is a center ID human rank hobby combined score calculation part and its periphery block diagram. It is a figure which shows the example of a score calculation parameter. It is a figure which shows the example of all user friend hobby storage DB. (A) is a figure which shows the example of a neighborhood friend matrix, (b) is a figure which shows the example of a social graph. (A) is a figure which shows the example of an initial rank vector, (b) is a figure which shows the example of a personalized vector. It is a figure which shows the example of center ID human rank storage DB. It is a figure which shows the example of hobby vector storage DB. It is a figure which shows the example of human relationship score storage DB.

  Hereinafter, modes for carrying out the present invention (hereinafter referred to as “embodiments”) will be described in the order of outline, configuration, and processing with reference to the drawings.

(Overview)
First, an outline of the important person search apparatus of this embodiment will be described. In the important person search apparatus of the present embodiment, human relationship information on the network (for example, friend list, blog, comment, etc. in SNS) and personal hobby information (for example, profile information in SNS, community affiliation information, etc.) are obtained. Use to search for important people for any person.

  At that time, the human relationship is converted into a social graph (in mathematical terms, expressed as an adjacency matrix or a transition matrix (transposition matrix of the adjacency matrix)), and the hobby information is vectorized. And by calculating such information according to the Markov process, it is possible to search for an important person for an arbitrary person in consideration of an indirect friendship such as a friend of a friend. The Markov process is a stochastic process having the property that the future behavior is determined only by the current value and is not related to the past behavior.

  Also, when creating a social graph (transition matrix (equivalent to the adjacency matrix in this embodiment)), only a person that is important to the user (subject) (for example, within 3 hops from the user (that is, a friend of a friend) Extract))). Then, a transition matrix for Markov process calculation is created using only the extracted important person candidates. As a result, the number of dimensions of the transition matrix for Markov process calculation can be greatly reduced, and the amount of calculation can be greatly reduced by performing personalized rank calculation using the created transition matrix with a small dimension. .

(Constitution)
Next, the configuration of the important person search apparatus of this embodiment will be described. As shown in FIG. 1, the important person search device 100 is a computer device, and includes an input unit 105, a user ID acquisition unit 110, a human relationship score calculation unit 120, a human relationship score acquisition unit 140, a human relationship score storage DB 150 ( Storage section) and result display section 160.

  The input unit 105 is a means for inputting information, and includes, for example, a keyboard, a mouse, a disk drive device, an input interface, and the like.

  The user ID acquisition unit 110, the human relationship score calculation unit 120, and the human relationship score acquisition unit 140 are, for example, a control device including a CPU (Central Processing Unit) and a RAM (Random Access Memory) (details will be described later). .

  The human relation score storage DB 150 is a means for storing a human relation score (a value indicating the importance of a certain person (other person) with respect to a certain person (subject), details will be described later), for example, from a general hard disk device or the like. Composed. In addition to the human relationship score storage DB 150, the important person search device 100 includes a storage unit that stores an operation program and various data, but illustration and detailed description thereof will be omitted or described later.

  The result display unit 160 is a means for displaying a calculation result (a list of important persons, etc.), and is a liquid crystal monitor, for example.

  The human relationship score calculation unit 120 is connected to all user friend hobby storage DBs 130 (storage units) outside the important person search device 100. The all-user friend hobby storage DB 130 is a storage unit in a server on the Internet, for example, and stores friend information (personal relationship information) and hobby information of all users of the SNS.

(processing)
Next, processing of the important person search apparatus 100 will be described with reference to FIGS. 2 to 16 (refer to other figures as appropriate). Here, a system that presents a list of important persons for a certain user to the user in the SNS is considered as an example. The important person search device 100 performs a pre-calculation process (a process before an important person search) performed in advance and a human relation score acquisition process (a process during an important person search) performed after the pre-calculation process is completed. Can be divided roughly.

  Note that the step numbers in FIG. 2 (for example, “S105” and “105” in step S110) are the same as the codes in the configuration in FIG. 1 (for example, “105” in the input unit 105 and user ID acquisition unit 110, “110”). Hereinafter, the outline of the process and the details of the process will be described separately.

(Outline of processing)
First, an outline of processing of the important person search apparatus 100 will be described.
As shown in FIG. 2, in the pre-calculation process, the input unit 105 accepts input of score calculation parameters (parameters used in formula (1) and formula (2) described later, details will be described later) by the user (steps). S105).

  Next, the human relation score calculation unit 120 acquires all user's friend information and hobby information from the all user friend hobby storage DB 130, and further uses all score calculation parameters for all users participating in the SNS. A human relation score is calculated for the human relation score, and the human relation score is stored in the human relation score storage DB 150.

Thereafter, when the user uses the important person search apparatus 100, first, the user ID acquisition unit 110 acquires the user ID input by the user via the input unit 105 (step S110).
Next, the human relationship score acquisition unit 140 acquires a human relationship score corresponding to the user ID from the human relationship score storage DB 150 (step S140).

  Subsequently, the result display unit 160 displays the result (human relation score). By viewing this display, the user can know the important person for himself (details will be described later).

(Details of processing)
Next, details of the processing of the important person search apparatus 100 will be described.
First, the human relation score calculation unit 120 (see FIG. 1) will be described. As shown in FIG. 3, in the human relationship score calculation unit 120, the score calculation parameter acquisition unit 121 acquires a score calculation parameter. This score calculation parameter is illustrated in FIG. Of the score calculation parameters shown in FIG. 10, α and γ are constants used in equations (1) and (2) described later, respectively, and are predetermined by the user as numerical values greater than 0 and smaller than 1 (details). Is described later). The number of acquired friend hops (= the number of Markov calculation repetitions) and the number of users with higher hobby scores are the constraint conditions at the time of calculation in the expressions (1) and (2) described later.

  Returning to FIG. 3, next, all user ID acquisition unit 122 acquires all user IDs from all user friend hobby storage DB 130. As illustrated in FIG. 11, the all-user friend hobby storage DB 130 has items of a user ID, a friend of the user, and a hobby of the user.

  Returning to FIG. 3, next, the center ID human rank calculation unit 200 (human rank calculation unit) acquires the score calculation parameters acquired from the score calculation parameter acquisition unit 121 and the users acquired from the all user ID acquisition unit 122. Using the friend information acquired from the all-user friend hobby storage DB 130 based on the ID, the central ID human rank (the reliability of the other person for the person (subject) of the central ID. Hereinafter, it is also simply referred to as “human rank”. .)

  Further, the central ID hobby vector creation unit 300 (common hobby information creation unit) uses all the scores based on the score calculation parameters acquired from the score calculation parameter acquisition unit 121 and the user IDs acquired from the all user ID acquisition unit 122. The central ID hobby vector (common hobby information. Information about the number of hobbies and the number of hobbies common to others with the central ID hobby vector (common hobby information). Vector (hereinafter also referred to as “hobby vector”, details will be described later).

  Thereafter, the center ID human rank hobby combined score calculation unit 400 (human relation score calculation unit) obtains the center ID human rank acquired from the center ID human rank calculation unit 200 and the center ID hobby acquired from the center ID hobby vector creation unit 300. Using the vector, the score calculation parameter acquired from the score calculation parameter acquisition unit 121, and the user ID acquired from the all user ID acquisition unit 122, the human relationship score is calculated, and the human relationship score is calculated as the human relationship score. Store in the storage DB 150.

  Next, the center ID human rank calculation unit 200 (see FIG. 3) will be described. As shown in FIG. 4, in the center ID human rank calculation unit 200, the center ID setting unit 210 is still subjected to the center ID human rank calculation processing among all the user IDs acquired from the all user ID acquisition unit 122. The user ID that has not been set is set as the central ID (at the start, for example, ID = 0 is set).

Next, the human rank calculation unit 220 acquires the score calculation parameter acquired from the score calculation parameter acquisition unit 121 and the all user friend hobby storage DB 130 for the user ID (center ID) set by the center ID setting unit 210. The human rank is calculated using the friend information of the user with this user ID.
The center ID human rank storage unit 230 stores the human rank calculated by the human rank calculation unit 220 in the center ID human rank storage DB 235 (storage unit).

  After the processing of the center ID human rank storage unit 230, the center ID setting user remaining determination unit 240 determines whether or not there are remaining users who are not set to the center ID. If the user remains, the process moves to the center ID setting unit 210. If the user does not remain, the process by the center ID human rank calculation unit 200 is terminated. The human rank information stored in the central ID human rank storage DB 235 is used later by the central ID human rank hobby combined score calculation unit 400.

  In the following, for example, a case where the user ID = 0 is set as the center ID in the center ID setting unit 210 will be considered.

  The human rank calculation unit 220 (see FIG. 4) will be described. As shown in FIG. 5, in the human rank calculation unit 220, the neighborhood friend matrix creation unit 221 includes friend information acquired from the all-user friend hobby storage DB 130 (specifically, directly or one or more with the central person and the central person). Information on a person indirectly having a human relationship via the person) is used to create a neighborhood friend matrix (person set matrix). It should be noted that the number of dimensions of this neighborhood friend matrix is the central person (one person) of the friend information and the person who has a personal relationship with the central person directly or indirectly through one or more persons (FIG. 12 (a ) In the example shown in FIG. 12 (a), there are 6 people in the example shown in FIG. 12A). Here, it is assumed that the created neighborhood friend matrix is as shown in FIG. 12A (details will be described later). In that case, the corresponding social graph is as shown in FIG. In FIG. 12B, a numerical value surrounded by a circle indicates a user ID, and a double arrow between the circles indicates that the persons with the user ID on both sides have a friendship.

  Returning to FIG. 5, the center ID initial rank vector creation unit 222 matches the number of dimensions (number of rows and columns) of the neighborhood friend matrix created by the neighborhood friend matrix creation unit 221, and sets the corresponding location of the center ID. An initial rank vector (an initial value of a rank vector; initial value as shown in FIG. 13A) set to “1” (a value other than “0”; the same applies hereinafter), and other portions are set to “0”. Vector). The rank vector is a vector used in the following formula (1), and a component is a vector indicating the human rank of each person. Similarly, the center ID personalized vector creation unit 223 matches the number of dimensions (n) of the neighborhood friend matrix created by the neighborhood friend matrix creation unit 221 and sets (n−1) / A personalized vector (as shown in FIG. 13B) in which n (a fixed value not less than “0.5” and less than “1”) is set to 1 / {n (n−1)} for the other elements. A vector that is created for each person who is the subject and indicates the importance of the person and others.

Returning to FIG. 5, the Markov calculation unit 224 uses the score calculation parameter acquired from the score calculation parameter acquisition unit 121 to set the initial rank vector of the center ID = 0 to r ₀ and the rank vector that has been subjected to the Markov process i times. R _i , a personalized vector v, and a neighborhood friend matrix A, a recurrence formula (predetermined recurrence) representing the Markov process for the number of times specified by the score calculation parameter “number of Markov computation iterations” (see FIG. 10) The following equation (1) is calculated.
r _{i + 1} = (1−α) Ar _i + αv (1)

  Note that the parameter α represents a ratio that the person (central person) set as the central ID trusts between himself and a person other than himself. Therefore, it shows that the ratio which trusts the third party other than oneself becomes large, so that (alpha) is small. That is, “1-α” is a first coefficient indicating the reliability of the person other than the central person from the central person. “Α” is a second coefficient indicating the reliability of the central person from the central person. In the present embodiment, the reliability and the importance are considered to have a positive correlation.

The expression (1), which is a recurrence formula, will be described more specifically. When the neighborhood friend matrix A is used to calculate Ar _i , the human rank as each component in the rank vector r _i Play a role to propagate to friends. This is because, as can be seen from FIG. 12A and FIG. 13A, when calculating Ar _i , among the components of r _i , “1” indicating that there is a friendship in the neighborhood friend matrix A This is because the part to be multiplied is reflected in the calculation result. This, for example, the first line of the neighboring friend matrix A used to calculate the components of the user ID = 0 in _{Ar i (0 1 1 1 0} 0) is taken to the user ID = 0, other than you It can also be understood from the fact that only the friend's part is “1”.

  Further, in the equation (1) that is a recurrence formula, v means that oneself is greatly trusted (trusted). This is because, as can be seen with reference to FIG. 13B, the personalized vector related to user ID = 0 has the numerical value “5/6” in the row of user ID = 0, It is because it is significantly larger than the numerical value of.

  As a specific example, α = 0.5 indicates that the person himself and the person other than himself are equally trusted. Further, for example, in the case of a network with a low reliability (human relations), the reliability of a person other than yourself can be lowered by increasing α.

In the above formula (1), specifically, r ₁ is calculated with r _i = r ₀ in the _first calculation, and calculation is performed with r _i = r ₁ in the second calculation. Similarly, the recurrence calculation according to the equation (1) is repeated. As described above, by using the initial rank vector and the personalized vector, in which only “1” is set for the person with the center ID, in the Markov process calculation, the higher the calculation result (the person with the center ID) Rank (human rank vector having human rank as a component).

  When repeating the recurrence calculation according to formula (1), the first calculation is to calculate the degree of belief in his direct friend, and the second calculation is to believe in his friend's friend. The degree is being calculated.

Thereafter, the Markov calculation unit 224 passes the calculation result according to the formula (1) to the central ID human rank storage unit 230 as the central ID human rank. Each component of the rank vector r _i corresponds to the center ID human rank. Further, the center ID human rank storage DB 235 (see FIG. 4), in which the center ID human rank storage unit 230 stores the center ID human rank, includes items of a center ID and a human rank, for example, as shown in FIG. The human rank item stores a human rank value for another person for each center ID. For example, for the person with the center ID = 0, the human rank of the user ID “1” is “0.1402777...” And the human rank of the user ID “2” is “0.1166.

  Next, the neighborhood friend matrix creation unit 221 (see FIG. 5) will be described. As shown in FIG. 6, in the neighborhood friend matrix creation unit 221, the friend ID acquisition unit 2211 acquires the friend information of the user ID of the center ID set by the center ID setting unit 210 from the all-user friend hobby storage DB 130. To do. Specifically, when the center ID = 0, the friend with the user ID = 0 is the user with the user ID = 1, 2, 3 (see FIG. 11). Since the process of acquiring this friend's user ID was performed once, the number of acquired friend hops = 1.

  Next, the neighborhood friend matrix storage unit 2212 adds the user with the user ID = 1, 2, 3 to the neighborhood friend matrix. That is, as shown in FIG. 12A, in the neighborhood friend matrix, the columns of ID = 1, 2, 3 in the row of ID = 0 are “1”, and the other columns are “0”. In the neighborhood friend matrix, “1” means a friend and “0” means not a friend. Further, the friends of ID = 1, 2, 3 are referred to as the friends of the number of hops = 1 from the person with the center ID = 0.

  Subsequently, the acquired friend hop count confirmation unit 2213 determines whether or not the specified acquired friend hop count (see FIG. 10) has been reached. In the case of a specific example, the number of friend hops acquired in the score calculation parameter = 2 (see FIG. 10), and the process by the friend ID acquisition unit 2211 here is one time. For this reason, this time, the processing by the friend ID acquisition unit 2211 and the neighborhood friend matrix storage unit 2212 is performed again using IDs 1, 2, and 3, which are the IDs of the acquired friends.

  Specifically, in the same manner as the process of acquiring the friend with the center ID = 0, the friend with the center ID = 1, the friend with the center ID = 2, and the friend with the center ID = 3 are acquired, and those friends Is added to the neighborhood friend matrix data as a friend with the number of hops = 2. That is, in the neighborhood friend matrix of FIG. 12, for example, the columns of ID = 0, 2, 4 in the row of ID = 1 are “1”, and the other columns are “0”. In addition, other rows are processed in the same manner, and the neighborhood friend matrix is processed so as to be a square matrix in which the rows and columns are symmetric (that is, the neighborhood friend matrix shown in FIG. 12 is completed).

  Since the acquired friend hop count confirmation unit 2213 has reached the acquired friend hop count = 2 (see FIG. 10), it ends the neighborhood friend matrix creation process, and the center ID initial rank vector creation unit 222 and the center ID personalized vector creation unit The number of vector space dimensions (“6” in the specific example) is transmitted to 223, and the neighborhood friend matrix is transferred to the Markov calculator 224.

  Next, the center ID hobby vector creation unit 300 (see FIG. 3) will be described. As shown in FIG. 7, in the center ID hobby vector creation unit 300, the center ID setting unit 310 has not yet performed the hobby vector creation process among all the user IDs acquired from the all user ID acquisition unit 122. The user ID is set as the center ID (at the start, for example, ID = 0 is set).

Next, the hobby vector creation unit 320 acquires the user ID (center ID) set by the center ID setting unit 210 from the score calculation parameters acquired from the score calculation parameter acquisition unit 121 and the all-user friend hobby storage DB 130. The hobby vector is created using the hobby information.
The hobby vector storage unit 330 stores the hobby vector created by the hobby vector creation unit 320 in the hobby vector storage DB 350 (storage unit).

  After the processing of the hobby vector storage unit 330, the center ID setting user remaining determination unit 340 determines whether there are any users who are not set to the center ID. If the user remains, the process moves to the center ID setting unit 310. If the user does not remain, the process by the center ID hobby vector creation unit 300 is terminated. Information in the hobby vector storage DB 350 is used later by the center ID human rank hobby combined score calculation unit 400.

  Next, the hobby vector creation unit 320 (see FIG. 7) will be described. As shown in FIG. 8, in the hobby vector creation unit 320, the center ID hobby information acquisition 321 acquires hobby information corresponding to the center ID input from the center ID setting unit 310 from the all-user friend hobby storage DB 130. In the case of a specific example, when the center ID = 0, “sports” and “cooking” are hobby information of the user ID = 0 (see FIG. 11).

  Next, the same hobby user information acquisition unit 322 acquires user information having the same hobby as the user with the center ID. When the hobby vector creation unit 320 acquires the hobby information from the all-user friend hobby storage DB 130 shown in FIG. 11, for example, except for the user ID = 0, only the user ID = 100 has “sport” as a hobby. In addition, a user having “cooking” as a hobby is only a user ID = 0.

  Next, the hobby vector generation unit 323 creates a hobby vector based on the information acquired from the central ID hobby information acquisition 321 and the same hobby user information acquisition unit 322 (details will be described later in FIG. 15).

  After the processing of the hobby vector generation unit 323, the all user processed confirmation unit 324 determines whether there is a user ID that has not been processed by the center ID hobby information acquisition 321. If the user ID exists, the process proceeds to the central ID hobby information acquisition 321. If the user ID does not exist, all the hobby vectors created by the hobby vector generation unit 323 are transferred to the hobby vector storage unit 330, and the processing of the hobby vector creation unit 320 ends.

  Next, the hobby vector storage DB 350 will be described. As shown in FIG. 15, the hobby vector storage DB 350 has items of center ID and hobby vector. Here, in the row of the center ID = 0 shown in FIG. 15, the user with the center ID = 0 has two hobbies, and only the user ID = 100 has one hobby as a common hobby (“sports”). It means that I had it. In this specific example, hobbies vectors that have been tabulated for both the “sports” and “cooking” hobbies are created. However, for example, it is possible to create a hobby vector for only “sports”.

  Next, the center ID human rank hobby combination score calculation unit 400 (see FIG. 3) will be described. As shown in FIG. 9, in the center ID human rank hobby combined score calculation unit 400, the calculation target ID setting unit 410 acquires a user ID that has not yet been subjected to human relationship calculation processing, and obtains the center ID (calculation target ID). ). Hereinafter, an example in which the user ID = 0 is described.

The center ID human rank acquisition unit 420 acquires the center ID human rank of the user ID = 0 from the center ID human rank storage DB 235.
The hobby vector acquisition unit 430 acquires the hobby vector of the user ID = 0 from the hobby vector storage DB 350.

  The hobby vector superior ID human rank acquisition unit 440 has the central ID human of the top k users out of the hobby vector elements (numerical values) in the hobby vector of user ID = 0 excluding the central ID. The rank is acquired from the center ID human rank storage DB 235. For example, in the example of FIG. 15, there is only one user (user ID = 100) excluding the central ID, but when there are a plurality of users, a hobby vector element (“1” for user ID = 100). The top k people except the central ID (user ID = 0) (in the specific example, “k = 1” because “the number of top hobby score users = 1” in the score calculation parameter of FIG. 10) The central ID human rank of the user is acquired from the central ID human rank storage DB 235.

The human relationship score calculation unit 450 calculates a human relationship score using the following formula (2) (predetermined human relationship score calculation formula). Here, for the user with user ID = p, the human relation score is S _p , the central ID human rank is r _p , and the size of the hobby vector element is the top k people (the top k people excluding the central ID element) ) Is the central ID human rank of the user corresponding to the element of q _i (i = 1, 2,..., K), and the element of the hobby vector is h _i (i = 1, 2,..., K). And
S _p = (1−γ) r _p + {γ × {Σ _{i = 1} ^k (h _i q _i )} / Σ _{i = 1} ^k h _i } (2)
(“Σ _{i = 1} ^k ” means the sum from i = 1 to i = k.)

  Note that the parameter γ represents the degree of importance placed on people who share a common hobby (preference). Therefore, if γ is large, it means that people with a common hobby are more important. That is, “1-γ” is a third coefficient indicating the importance level of the human rank. “Γ” is a fourth coefficient indicating the importance of the commonality of hobbies.

Further, “{Σ _{i = 1} ^k (h _i q _i )} / Σ _{i = 1} ^k h _i ” is a value indicating the degree of commonality of hobbies between the central person and all other persons. Regarding Formula (2), γ = 0.1 as a specific example means that the importance of a person with a common hobby is considered to be about 1/10 of that of his direct friend.

The human relationship score storage unit 460 stores the human relationship score calculated by the human relationship score calculation unit 450 in the human relationship score storage DB 150.
After the process of the human relationship score storage unit 460, the calculation target ID non-set user remaining determination unit 470 determines whether there are any users who are not set as the calculation target ID. If the user remains, the process proceeds to the calculation target ID setting unit 410. If the user does not remain, the process by the central ID human rank hobby combined score calculation unit 400 is terminated. Thereby, the human relation score can be calculated for all users, and the human relation score can be stored in the human relation score storage DB 150.

  Next, the human relation score storage DB 150 will be described. As shown in FIG. 16, the human relationship score storage DB 150 includes items of a user ID and a human relationship score. The user can know the important person for himself / herself by viewing the contents of the human relation score storage DB 150 displayed on the result display unit 160.

  In this manner, according to the important person search apparatus 100 of the present embodiment, friends of friends using friend information that is direct human relationship information and hobby information that is indirect human relationship information. By analyzing such graph structure information, it is possible to search for an important person for an arbitrary person while considering indirect human relationships.

  Also, when calculating the human relationship score, only the number of people within a few hops from the user (subject) is extracted, and a matrix or vector with a small number of dimensions is used in the calculation. It is possible to search for an important person for an arbitrary person. In other words, compared to the case of using a huge transition matrix for all users and a huge hobby vector having the same number of dimensions as the number of users, almost the same effect (important person search results) is obtained with a small amount of calculation. be able to.

  Further, by creating a program to be executed by a computer constituting the important person search apparatus 100 and installing the program in the computer, the computer can realize each function based on the program.

Although description of this embodiment is finished above, the aspect of the present invention is not limited to these.
For example, the all-user friend hobby storage DB 130 may exist inside the important person search apparatus 100, not necessarily outside the important person search apparatus 100.
In addition, specific configurations and processes can be appropriately changed without departing from the gist of the present invention.

DESCRIPTION OF SYMBOLS 100 Important person search apparatus 105 Input part 110 User ID acquisition part 120 Human relation score calculation part 130 All user friend hobby storage DB
140 Human relationship score acquisition unit 150 Human relationship score storage DB
160 Result display section 200 Center ID human rank calculation section (human rank calculation section)
300 Central ID hobby vector storage unit (common hobby information creation unit)
400 Center ID human rank hobby combined score calculator (human relation score calculator)

Claims (7)

An important person search device that searches for an important person for an arbitrary person from human relation information that indicates the presence or absence of a human relationship between any two persons and hobby information that indicates the hobbies of each person. A person search method,
The important person search device includes:
A storage unit for storing the human relation information and the hobby information;
Using the human relationship information and the hobby information stored in the storage unit, a human relationship score indicating the importance of another person for an arbitrary person is calculated, and the human relationship score is stored in the storage unit A human relationship score calculator,
A human relationship score acquisition unit that acquires, from the storage unit, a human relationship score of another person with respect to the person when receiving input of identification information of an arbitrary person;
A result display unit that displays the human relationship score acquired by the human relationship score acquisition unit,
In the human relationship score calculation unit,
The human rank calculator
A person who has a predetermined person as a central person and refers to the human relation information stored in the storage unit and who has a human relation directly with the central person or indirectly through one or more persons is referred to as “ Create a person set matrix with a value other than “0” and other people with “0”
A human rank indicating the importance of the central person based on the human relation information between the central person and each person who has a human relation directly with the central person or indirectly through one or more persons. Create a predetermined initial vector for the human rank vector
Creating a personalized vector whose constituent elements are the ratio of trustworthiness from the central person and each person who has a human relationship with the central person directly or indirectly through one or more persons,
The product of the person set matrix and the human rank vector is multiplied by a first coefficient indicating the reliability of the central person other than the central person, and the reliability of the central person from the central person is obtained. For the predetermined recurrence formula based on the Markov process of multiplying the personalized vector by the second coefficient shown and adding the two multiplied vectors to update the human rank vector, the person set matrix And applying the initial vector of the human rank vector and the personalized vector a predetermined number of times to calculate a human rank vector having the human rank as a component,
The common hobby information creation department
With reference to the hobby information stored in the storage unit, the common number of hobbies with each other person is calculated with respect to the central person, the number of hobbies with respect to the central person, and the common hobby with the central person Creating common hobby information indicating the number of common hobbies for each person having
Human rank hobby coupling part
For each of the other persons, a third coefficient indicating the importance level of the human rank is multiplied by the human rank of the other person in the calculated human rank vector to indicate the importance level of the hobby commonness. The degree of commonality of the central person and the central person directly or indirectly through one or more persons is calculated based on the common hobby information. Calculated by applying the human rank and the common hobby information to a predetermined human relationship score calculation formula that multiplies the value and adds the two multiplied values to the human relationship score. Then, a human relation score of the other person with respect to the central person is calculated.   The important person search method according to claim 1, wherein the person set matrix is an adjacency matrix or a transition matrix.   The human relation score calculation unit applies a predetermined number from the larger number of the common hobbies in the hobby vector to the predetermined human relation score calculation formula. Important person search method. An important person search device that searches for an important person for an arbitrary person from human relation information indicating the presence or absence of a human relation between two arbitrary persons and hobby information indicating a hobby of each person. And
A storage unit for storing the human relation information and the hobby information;
Using the human relationship information and the hobby information stored in the storage unit, a human relationship score indicating the importance of another person for an arbitrary person is calculated, and the human relationship score is stored in the storage unit A human relationship score calculator,
A human relationship score acquisition unit that acquires, from the storage unit, a human relationship score of another person with respect to the person when receiving input of identification information of an arbitrary person;
A result display unit that displays the human relationship score acquired by the human relationship score acquisition unit,
The human relation score calculation unit
A person who has a predetermined person as a central person and refers to the human relation information stored in the storage unit and who has a human relation directly with the central person or indirectly through one or more persons is referred to as “ Create a person set matrix with a value other than “0” and other people with “0”
A human rank indicating the importance of the central person based on the human relation information between the central person and each person who has a human relation directly with the central person or indirectly through one or more persons. Create a predetermined initial vector for the human rank vector
Creating a personalized vector whose constituent elements are the ratio of trustworthiness from the central person and each person who has a human relationship with the central person directly or indirectly through one or more persons,
The product of the person set matrix and the human rank vector is multiplied by a first coefficient indicating the reliability of the central person other than the central person, and the reliability of the central person from the central person is obtained. For the predetermined recurrence formula based on the Markov process of multiplying the personalized vector by the second coefficient shown and adding the two multiplied vectors to update the human rank vector, the person set matrix A human rank calculation unit that calculates a human rank vector having the human rank as a constituent element by calculating a predetermined number of times by applying the initial vector of the human rank vector and the personalized vector;
With reference to the hobby information stored in the storage unit, the common number of hobbies with each other person is calculated with respect to the central person, the number of hobbies with respect to the central person, and the common hobby with the central person A common hobby information creating unit that creates common hobby information indicating the number of common hobbies for each person having
For each other person, a third coefficient indicating the importance level of the human rank is multiplied by the human rank of the other person in the calculated human rank vector to indicate the importance level of the hobby commonness. The degree of commonality of the central person and the central person directly or indirectly through one or more persons is calculated based on the common hobby information. Calculated by applying the human rank and the common hobby information to a predetermined human relationship score calculation formula that multiplies the value and adds the two multiplied values to the human relationship score. A human rank hobby combining unit that calculates a human relation score of the other person with respect to the central person,
An important person search device comprising:   The important person search apparatus according to claim 4, wherein the person set matrix is an adjacency matrix or a transition matrix.   5. The human relation score calculation unit applies a predetermined number from a larger number of the common hobbies in the hobby vector to the predetermined human relation score calculation formula. Important person search device. The program for functioning a computer as an important person search device of any one of Claims 4-6.

Images